Final Year Project — Precision Farming Robot 2.0

Project Overview

The Precision Farming Robot 2.0 is an autonomous agricultural robotics system designed for smart farming and precision field operations. It integrates embedded systems, ROS2 control architecture, computer vision, and remote monitoring platforms into a single unified system.

The robot performs advanced agricultural tasks such as weed detection, crop field scanning, environment mapping, and targeted spraying. It uses AI-driven perception models combined with real-time sensor fusion from IMU, encoders, and camera streams to make intelligent navigation decisions.

This system is designed as a complete robotics pipeline where low-level control, high-level decision-making, and user interaction layers all work in synchronization. It represents a real-world deployment of ROS2 distributed robotics architecture integrated with embedded hardware and deep learning models.

This project was developed as a Final Year Project (FYP) to demonstrate industrial-grade robotics engineering, combining AI, control systems, and IoT communication.

System Architecture

The system follows a multi-layer distributed architecture where each layer is responsible for a specific function: embedded firmware handles real-time motor and sensor control, Raspberry Pi runs ROS2 for coordination, and AI nodes process visual data for decision-making.

Communication between modules is handled using ROS2 topics and services, ensuring low-latency and scalable data exchange. MQTT is used to bridge robot data to external applications such as mobile and web dashboards.

This architecture allows the system to remain modular, fault-tolerant, and easily extendable for future agricultural automation tasks.

Architecture Diagram

Full ROS2 + Embedded + AI System Design

Robot Hardware System

The robot is built on a 4-wheel differential drive platform designed for stability on uneven agricultural terrain. It uses a Raspberry Pi 4 as the main processing unit, supported by Arduino Uno and Nano microcontrollers for real-time control tasks.

Motor actuation is handled through L298N drivers, while encoders provide feedback for precise odometry. An IMU sensor is used for orientation tracking, and a camera module captures real-time field images for AI processing.

This separation of responsibilities ensures deterministic control at the firmware level while keeping high-level computation on ROS2.

Hardware System

Mechanical + Embedded Integration

AI Weed Detection System

The vision system is powered by a YOLO-based deep learning model trained on agricultural datasets for weed identification. It processes live camera frames and detects weeds in real time with bounding box localization.

The detected results are published to ROS2 topics, allowing the robot to take intelligent actions such as navigation adjustments and targeted spraying based on detection confidence and position.

This system enables real-time perception and decision-making for precision agriculture without relying on marker-based localization.

AI Weed Detection

YOLO + ArUco Vision Pipeline

Desktop Control Panel (Digital Twin)

The desktop application is a real-time digital twin system built using Qt6 C++. It visualizes robot state, camera feed, sensor data, and ROS2 topics in a unified interface.

It also provides manual override controls, allowing operators to directly send velocity commands, monitor telemetry, and debug robot behavior in real time.

The system supports multiple modes including live robot mode, simulation mode, and offline analysis mode, making it suitable for both testing and deployment.

Desktop Control System

Real-Time ROS2 Digital Twin

Mobile Monitoring System

The mobile application enables remote monitoring of robot operations using MQTT-based communication. It provides live telemetry updates, detection results, and remote command capabilities.

This allows field operators to interact with the robot even without direct access to the desktop system, improving mobility and usability in real agricultural environments.

Mobile Monitoring System

MQTT-Based Remote Control

ROS2 Control System

The ROS2 layer acts as the central communication backbone of the system. It manages motion control, sensor fusion, odometry, camera streaming, and AI integration.

Each function is implemented as an independent node, enabling modular development and parallel execution. This ensures system scalability and reliability under real-time constraints.

ROS2 Control System

Distributed Robotics Framework

Key Features

• Autonomous Navigation: ROS2-based differential drive control with encoder feedback.
• AI Vision System: Real-time YOLO detection for weed identification and classification.
• Sensor Fusion: Combines IMU, encoders, and camera data for accurate positioning.
• Digital Twin Interface: Real-time desktop visualization and debugging system.
• Mobile Control System: MQTT-based remote monitoring and command execution.
• Distributed Robotics Design: Embedded + ROS2 + AI layered architecture.

Technologies Used

This project demonstrates an end-to-end robotics pipeline integrating embedded systems, AI perception, and distributed ROS2 architecture for real-world agricultural automation.

DevNest

DevNest — A Unified Desktop Control Panel for Developers & Self-Hosters

Overview

DevNest is a high-performance desktop application that centralizes management of Docker containers, Tailscale nodes, SSH connections, system services, logs, and files into a single, customizable interface. It is built with React 19, TypeScript, and a Rust (Tauri) backend to deliver native-level performance and secure OS integrations.

Problem It Solves

Developers and homelab operators often switch between many CLIs and dashboards, losing context and risking credential exposure. DevNest reduces friction by providing a unified, resilient, and secure dashboard for both local and remote machine management.

Key Features

• Unified Infrastructure Management: Control Docker, Tailscale, system services, and files from a single UI across local and remote hosts.
• Advanced Workspace & Pane System: A Tmux-like binary-tree window manager for splitting and organizing panes.
• Resilient Remote Connections: Heartbeat monitoring with automatic reconnection using exponential backoff.
• Developer Tools: Built-in terminal, log viewer, and a lazy-loaded SQL client powered by the Monaco editor.
• Command Palette & Shortcuts: VS Code-style command palette and customizable global keyboard shortcuts.
• Secure Credential Storage: Native OS Keyring integration via Rust to avoid storing secrets in plain text.

Demo

Engineering Highlights

• Type-Safe IPC: Strict API layer in src/lib/api.ts ensuring end-to-end type safety between Rust/Tauri and the React frontend.
• Optimized Rendering & Lazy Loading: Workspace pane state implemented as pure functions over a binary tree; heavy components are lazy-loaded with `React.lazy`/`Suspense`.
• Modular State Architecture: Decoupled Zustand stores (`app-store`, `ui-store`, `sql-store`) with a rule that stores never call Rust directly.
• Native Performance: Rust backend handles SQLite persistence, SSH networking, keyring access, and device polling so the UI remains responsive.

Technologies Used

IoT-Based Water Tanker Management System

Project Overview

The IoT-Based Water Tanker Management System is an innovative solution designed to automate water monitoring, tanker ordering, and delivery management. The project integrates IoT devices, cloud computing, and mobile technology to create a smart and efficient water management ecosystem. The system continuously measures water level, temperature, and humidity using sensors like the Ultrasonic Sensor and DHT11, ensuring accurate environmental monitoring.

All collected data is sent to AWS IoT Core, where it is processed and stored in a secure cloud database. A React Native mobile application allows users to monitor their water tank in real-time and place tanker orders when the water level is low. Once an order is placed, the React-based admin portal receives it, assigns it to an available driver, and tracks delivery progress. The driver application (also built with React Native) helps drivers receive delivery instructions, confirm completion, and keep records synchronized with the cloud. The backend is powered by Node.js.

This system eliminates the need for manual monitoring and reduces communication gaps between users, administrators, and delivery teams. It serves as a scalable, real-world example of how IoT and cloud technology can improve daily utilities and resource management.

Mobile App for Customers

Real-Time Water Tank Monitoring

Mobile App for Drivers

Real-Time Order Tracking

Key Features

• Real-Time Monitoring: Continuously tracks the water level, temperature, and humidity using IoT sensors.
• Cloud Connectivity: Utilizes AWS IoT Core and MQTT protocol for seamless data transfer to the backend.
• User-Friendly Mobile App: Displays current tank water level and environmental data, and allows users to order tankers instantly.
• Admin Management Panel: Provides complete control for administrators to view all orders, assign drivers, and monitor deliveries.
• Driver Application: Helps drivers receive delivery details, navigate to locations, and update delivery status in real time.
• Database Integration: All sensor data, orders, and delivery records are stored in a structured and secure database.
• Notifications: Alerts users when the water level drops below a set threshold, suggesting when to order a tanker.

Technologies Used

How It Works

1. Sensing: The ESP32 reads data from the Ultrasonic Sensor (for water level) and DHT11 (for temperature and humidity).
2. Communication: The IoT device sends this data to AWS IoT Core using MQTT protocol.
3. Data Storage: The backend receives the data and stores it in a MongoDB database for future analysis.
4. User Interface: The mobile app fetches this data from the backend, displaying real-time tank levels and environmental readings.
5. Order Placement: When the water level is low, the user can place a tanker order directly from the app.
6. Admin Control: The admin portal receives the order, checks driver availability, and assigns the delivery task.
7. Delivery Execution: The driver app displays the assigned order, delivery address, and allows status updates (e.g., on the way, delivered).
8. Data Synchronization: All updates are synced in real-time across cloud, admin, and user dashboards.

Web Dashboard for Admin

Real-Time Order Management

Benefits

• Automation: Eliminates manual tank checks and phone-based ordering.
• Efficiency: Streamlines order management and delivery scheduling.
• Transparency: Keeps users, admins, and drivers connected in one system.
• Data Insights: Enables analysis of water usage trends and delivery patterns.
• Scalability: Supports multiple devices and users for large-scale deployments.
• Sustainability: Encourages smart water usage and reduces wastage.

This system demonstrates the power of IoT and cloud technology in solving real-world problems. By automating water management, it not only saves time and effort but also promotes efficient resource utilization and sustainability.

Precision Farming Robot

Project Overview

The Precision Farming Robot project explores the use of automation for weed management in farming. As an early prototype, it demonstrates the basics of weed detection and targeting. The project aims to reduce labor and herbicide usage, while laying the groundwork for more advanced agricultural robotics. The robot uses a camera to capture images of the surrounding environment, and basic image processing techniques to identify weeds. A laser pointer is then used to target the detected weeds, simulating weed elimination. The robot's movement is controlled via a desktop application, and data is exchanged through the MQTT protocol. This project highlights the potential of automation in agriculture, and future developments will focus on integrating AI for better weed detection, improving navigation, and adding more advanced sensors.

Precision Farming Robot

Automated Weed Detection and Targeting

Key Features

• Automated Weed Detection: Uses basic image processing techniques to detect weeds in a controlled environment.
• Targeting Mechanism: A laser pointer targets detected weeds for simulation of weed elimination.
• Controlled Navigation: The robot's movement is controlled via a desktop application within a limited environment.
• Modular Design: Easy to update and extend with new features, such as AI or better navigation.
• Decentralized Control: Independent control of each module, simplifying the design and development process.

Technologies

Benefits

• Weed Detection: Introduces automated weed detection as a proof of concept for more efficient farming methods.
• Labor Reduction: Aims to reduce manual labor for tasks like weed monitoring and elimination.
• Herbicide Reduction: By targeting weeds directly, the robot can reduce the need for herbicides, promoting sustainable farming practices.

Precision Farming Robot Application

Rover Controller Application

How It Works

The robot uses a camera to capture images of the surrounding environment. Basic image processing identifies weeds, and the robot moves to target them. The laser pointer is controlled by a servo motor for precise targeting. The robot is controlled via a desktop application, and data is exchanged through the MQTT protocol. This early prototype focuses on demonstrating the potential of automation in weed management and will be developed further in future versions.

Precision Farming Robot Architecture

Automated Weed Detection and Targeting

Challenges & Future Plans

• AI Integration: Future development will incorporate machine learning algorithms to enhance weed detection accuracy. The goal is to enable the robot to distinguish between weeds and crops in complex environments and automatically adjust its targeting strategy based on factors such as plant size and growth stage.
• Improved Navigation: Future versions will integrate GPS and advanced path-planning algorithms, allowing the robot to autonomously navigate large agricultural fields. Obstacle detection and avoidance systems will be implemented to ensure efficient operation in challenging terrains, providing a robust solution for field coverage.
• Scalability: The modular design will be further developed to handle larger systems for industrial-scale farms. The aim is to build a fleet of smaller robots that can work together autonomously, providing an efficient solution for larger farms while reducing the need for human supervision.
• Advanced Sensors: Future iterations will include additional sensors, such as thermal cameras or hyperspectral imaging, to enhance weed detection under various lighting conditions and improve precision in identifying specific types of weeds or other potential crop threats.
• Data Analytics: The robot will be integrated with cloud-based systems to collect and analyze data on weed density, soil conditions, and crop health. This data can be used to generate reports and insights for farmers, enabling better decision-making and fostering data-driven farming practices.
• Sustainability: Renewable energy sources, such as solar panels, will be incorporated to make the robot more energy-efficient and sustainable, minimizing the carbon footprint of its operations.
• Collaboration with Farmers: Future work will involve close collaboration with farmers to fine-tune the robot’s design and functionality. Gathering feedback from real-world users will be crucial to ensure the robot is adaptable and can meet the unique challenges of different types of farms.

This first prototype is just the beginning, and the development of new features is underway. With future improvements, the robot has the potential to revolutionize weed management in agriculture, making farming more sustainable and efficient.

FlashMind

Project Overview

FlashMind is an AI-powered learning platform built with Next.js, Firebase, OpenRouter (LLM API), Tello, and Clerk for authentication. It transforms learning through smart flashcards, personalized quizzes, and concept understanding using AI. Designed for both casual learners and serious students, it combines productivity tools with active recall strategies to enhance memory and performance. Users can create, edit, and organize flashcards, while AI suggests topics, adjusts difficulty, and generates quizzes based on weak areas.

FlashMind AI-Powered Learning

Smart Learning with AI-Generated Flashcards

Key Features

• AI-Generated Flashcards: AI creates flashcards from uploaded study materials, perfect for quick review and memorization.
• Study Streaks: Track your progress with daily streaks and long-term learning achievements.
• AI-Generated Quizzes: Personalized quizzes are created based on your weaknesses to strengthen your knowledge.
• Download Cards: Allow user to download Flashcards, allowing for flexible learning on the go.
• AI Recommendations: Get suggestions for topics to study based on your past performance and activity.
• Collaborative Flashcards: Share and collaborate on flashcards with others to enhance group learning.
• Dark Mode: Enable dark mode for a more comfortable study experience during late-night sessions.
• AI Chatbot Assistance: Get instant help from the AI chatbot to clarify concepts and boost learning efficiency.

Technologies

How It Works

FlashMind allows users to paste study materials such as notes or textbooks, which the AI analyzes to create personalized flashcards and quizzes. The AI tracks your performance, schedules spaced repetition reviews, and adjusts the difficulty based on how well you’ve mastered the content. Users can also share flashcards with others, collaborate on study sessions, and keep motivated with daily streaks. Authentication is powered by Clerk, ensuring secure access and personalized profiles.

FlashMind Application

AI-Enhanced Learning Experience

IoT Plant Monitoring System

Project Overview

I'm excited to share my IoT Plant Monitoring System—a project born from my passion for technology and plant care. Using an ESP32 microcontroller, this system tracks key health indicators like humidity, temperature, and soil moisture, and sends the data to the cloud via AWS IoT Core. It's all about using tech to make plant care smarter and easier!

Key Features

• Real-Time Monitoring: Stay updated on your plant's health around the clock.
• Cloud Integration: Access your data from anywhere, anytime, with AWS IoT Core.
• Accurate Time Sync: Precise data logging with NTP server synchronization.
• Precise Data Collection: Using DHT11 and soil moisture sensors for accurate readings.

Technologies

Benefits

• Smart Plant Care: Real-time insights for keeping your plants healthy and happy.
• Remote Access: Monitor your plants wherever you are.
• Reliable Records: Accurate, time-stamped data for peace of mind.

How It Works

• Data Collection: Sensors gather real-time data on plant conditions.
• Data Transmission: Sends data to AWS IoT Core using MQTT for remote monitoring.
• User Access: Check plant conditions from any device, anywhere you are.

IoT Plant Monitoring System

It Sends Data to AWS IoT Core using MQTT protocol.

Challenges & Opportunities

• Client Application: Building a user-friendly app to make monitoring even easier. Your input and ideas are highly appreciated!

Join Me!

• Custom PCB: Looking for help to design an integrated PCB.
• More Sensors: Let's explore adding more sensors for deeper insights.
• 3D Printed Case: Interested in creating a portable, sleek 3D-printed case? Let's collaborate!

If you’re passionate about IoT or just love plants, I’d love to have you join me on this journey. Let’s build something awesome together!

Battery Degradation Trajectory Prediction

Introduction

A dynamic project centered around the in-depth analysis of time-series data, specifically focusing on battery performance. Leveraging Python's powerful libraries, including pandas andmatplotlib, this project aimed to unravel critical insights fromCSV files, shedding light on the intricate behavior of various battery parameters over time.

My Role

1. Conceptualized the data analysis approach.
2. Implemented data loading, refinement, andmanipulation strategies.
3. Developed customplotting functions and threshold-based analysis.
4. Conducted capacity and voltage charging analyses, applying domain specific knowledge.
5. Oversaw datamerging and filtering processes for comprehensive insights.

Technologies

Battery Degradation Trajectory Prediction

Data-Driven Lifespan: Python Predictions for Battery Health

Open Inventory

Introduction

Using the Qt framework and C++, Open Inventory is an open-source inventory management system. The system offers a complete solution for tracking and managing assets, to streamline and improve inventory management for enterprises.

My Role

1. Custom menus for administrators, vendors, and customers provide an intuitive interface.
2. Order Placement Made Easy: Enables users to place orders and generate invoices with ease.
3. Effective Inventory Management: Provides a dedicated interface for suppliers, making it easier to add stock.
4. Strong Control: User Administration: Effective Instruments for Managing Vendor and Customer Profiles.
5. Financial analytics: Sturdy instruments that offer information on profit and loss to help with decision-making.

Technologies

Open Inventry

Open sourse inventry system.

Hostel Management System

Introduction

The Hostel Management System is a comprehensive solution for managing hostel operations. It offers features for hostel booking, maintenance requests, user administration, and more. Developed in Java with an MVC architecture, it aims to streamline hostel management for administrators, students, and staff.

My Role

1. Custom menus for administrators, maintenance staff, and students to ensure smooth system operation.
2. Streamlined order placement for students to book hostels easily.
3. Developed a user management system for handling administrator, staff, and student profiles effectively.

Technologies

Hostel Management System

Efficient management of hostel bookings and operations.